Autoanalyzer for Milk Quality Control Based on the Lactose, Fat, and

A novel autoanalyzer was developed to assess the quality of milk samples according to the percentage of lactose, fat, and total protein they contain. ...
0 downloads 0 Views 86KB Size
Anal. Chem. 2003, 75, 1425-1429

Autoanalyzer for Milk Quality Control Based on the Lactose, Fat, and Total Protein Contents Rafael Lucena, Mercedes Gallego, Soledad Ca´rdenas, and Miguel Valca´rcel* Department of Analytical Chemistry, Campus de Rabanales, University of Co´ rdoba, E-14071 Co´ rdoba, Spain

A novel autoanalyzer was developed to assess the quality of milk samples according to the percentage of lactose, fat, and total protein they contain. The module comprises two pumps (one of high pressure), an injection valve, a filter, and an evaporative light-scattering detector. A volume of 15 µL of dilute milk was injected in an ethanolwater (50% v/v) stream for precipitation/retention of protein/fat, being the lactose content determined in the filtrate. The fat fraction was calculated using an ethanol stream, and total protein was finally dissolved by means of a 1.7 mol/L acetic acid solution. The simplicity of the proposed automatic module lies in the universal response of the detector, which permits the determination of the three macrocomponents in milk. In addition, the flow injection method allows their sequential analysis in the same injected sample by using selective reagents for each compound. The proposed method was validated with an SRM milk sample as well as by comparison of the results obtained with those provided by the IR method. In addition, the proposed analyzer is cheaper than its counterpart that is based on infrared technique. Milk is an essential component of the human diet. In several countries, the milk payment in the industry is established according to its quality,1 which has improved the production because milk producers are now more interested in milk quality control. Taking into account that the industrial process cannot improve the milk quality, the raw material determines the quality of final product. Several parameters (such as percentage of lactose, fat, and protein as well as bacterial content and somatic cells) are used as reference values to establish the milk quality. 2 In addition, each country can set the level of these reference values, which can be further modified by the dairy industries, being the price directly connected with the values of these parameters. The individual determination of fat in milk is carried out using different approaches, namely, extraction with an organic solvent followed by a gravimetric determination using a piezoelectric detector,3 NMR,4 and NIR5 detection and the use of chromatographic techniques, such as gas (GC)6 or liquid chromatography * Corresponding author. Phone/fax: +34-957-218-616. E-mail: qa1meobj@ uco.es. (1) Varnan, A. H.; Sutherland, J. P. Tecnologı´a, quı´mica y microbiologı´a; Editorial Acribia: Zaragoza, Spain, 1995. (2) Reglamento No. 2597/97 del Consejo, 18 de diciembre de 1997, Diario Oficial de las Comunidades Europeas. (3) Manganiello, L.; Rı´os, A.; Valca´rcel, M.; Ligero, A.; Tena, T. Anal. Chim. Acta 2000, 406, 309. (4) Le-Botlan, D. J.; Helie, I. Analusis 1994, 22, 108. (5) Chen, J. Y.; Iyo, C.; Kawano, S. J. Near Infrared Spectrosc. 1999, 7, 265. 10.1021/ac020553n CCC: $25.00 Published on Web 02/13/2003

© 2003 American Chemical Society

(LC).7 Supercritical fluid extraction with different detectors8 has also been proposed. Liquid chromatography 9 is the alternative of choice for the individual determination of carbohydrates in milk. Amperometric biosensors,10 capillary electrophoresis,11 gas chromatography,12 and flow systems with UV-vis13 detection have also been referenced. The most important approaches to protein determination in milk are the capillary electrophoresis14 and liquid chromatography.15 The simultaneous determination of carbohydrates, fat, and casein in milk is usually carried out by IR spectroscopy.16-18 In this regard, different analyzers have been developed and are now commercially available, although they are rather expensive. The evaporative light-scattering detector (ELSD)19 measures the light scattered by the solutes present in the sample after evaporation of the mobile phase. The ELSD is a quasi-universal detector because only compounds less volatile than the carrier stream are detected. It is useful with a previous separation technique such as LC.20-22 In this work, the combined use of a continuous flow system and an ELSD for sequential determination of the main components of cow milk is described for the first time. The method is based on the precipitation of proteins with ethanolwater, with the fat also being retained as the likely result of its interaction with the molecules of casein. Lactose can be easily determined in the filtrate. By using selective solvents, the fat is first dissolved, and then total protein is also dissolved with dilute acetic acid. A single sample injection is required to determine the three parameters, and therefore, milk quality control is performed within a few minutes. (6) Feifel, S.; Pendl, R.; Caviezel, R. Chem. Lab. Biotechnol. 2000, 51, 48. (7) Spanos, G. A.; Schwartz, S. J.; van-Breemen, R. B.; Huang, C. H. Lipids 1995, 30, 85. (8) Dionisi, F.; Hug, B.; Aeschlimann, J. M.; Houllemar, A. J. Food Sci. 1999, 64, 612. (9) Ca´ceres, A.; Ca´rdenas, S.; Gallego, M.; Rodrı´guez, A.; Valca´rcel, M. Chromatographia 2000, 52, 314. (10) Adanyi, N.; Szabo, E. E.; Varadi, M. Eur. Food Res. Technol. 1999, 209, 220. (11) Lee, Y. H.; Lin, T. I. J. Chromatogr., B 1996, 681, 87. (12) Olano, A.; Calvo, M. M.; Reglero, G. Chromatographia 1986, 21, 538. (13) Yokoi, Y.; Matsubara, C.; Takamura, K. Bunseki Kagaku 1995, 44, 355. (14) Strickland, M.; Johnson, M. E.; Broadbent, J. R. Electrophoresis 2001, 22, 1510. (15) Guillou, H.; Miranda, G.; Pelissier, J. P. Lait 1987, 67, 135. (16) Dı´az-Carrillo, E.; Mun ˜oz-Serrano, A.; Alonso-Moraga, A.; Serradilla-Manrique, J. M. J. Near Infrared Spectrosc. 1993, 1, 141. (17) Lefier, D.; Grappin, R.; Pochet, S. J. AOAC Int. 1996, 79, 711. (18) Sasic, S.; Ozaki, Y. Anal. Chem. 2001, 73, 2294. (19) Kohler, M.; Haerdi, W.; Christen, P.; Veuthey, J. L. Trends Anal. Chem. 1997, 16, 475. (20) Stolyhwo, A.; Colin, H.; Guiochon, G. J. Chromatogr., A 1983, 288, 253. (21) Lafosse, M.; Elfakir, C.; Morin-Allorys, L.; Dreux, M. J. High. Resolut. Chromatogr. 1992, 15, 312. (22) Ca´rdenas, S.; Gallego, M.; Valca´rcel, M. Anal. Chim. Acta 1999, 402, 1.

Analytical Chemistry, Vol. 75, No. 6, March 15, 2003 1425

EXPERIMENTAL SECTION Apparatus. The experiments were carried out using a HewlettPackard 1050 high-pressure quaternary gradient pump for delivery of the reagent solutions and a six-port LC injection valve (Knauer 6332000) fitted with a 15-µL PTFE sample loop and a DDL 31 evaporative light scattering detector (Eurosep, Cergy-Pontoise, France) using air as nebulizing gas at 1.7 bar. The photomultiplier gain was set at 500 and 600 V (depending on the analyte), and the instrument temperature was fixed at 65 °C. The low-pressure flow system designed for sample aspiration consisted of a Gilson Minipuls-3 peristaltic pump fitted with poly(vinyl chloride) pumping tubes, PTFE tubing of 0.5-mm i.d for coils, and standard connectors. A laboratory-made filtration column (2.5 cm × 3 mm i.d) packed with cotton (∼50 mg) was also used. Signals were registered on a Radiometer (Copenhagen, Denmark) REC-80 Servograph recorder. Special attention should be paid to the dry residue of the solvent because it can increase the background noise, thus decreasing the sensitivity and affecting the performance of the instrument because of their deposit in the optical cells.19,23 For method validation, a MilkoScan FT 120 instrument (FOSS electric, Barcelona, Spain) was also used. Reagents and Samples. All reagents were of analytical grade or better. HPLC gradient grade ethanol supplied by Scharlau (Barcelona, Spain); acetic acid from Merck (Darmstadt, Germany) and Milli-Q ultrapure water (Millipore Corp, Madrid, Spain) were employed. Solutions of 1.7 mol/L acetic acid and ethanol-water (50% v/v) were prepared and filtered through a nylon 66 filter (0.45-µm pore size) before use. SRM infant formula no. 1846, with certified fat, protein, and carbohydrate concentrations, supplied by the National Institute of Standards & Technology (NIST, Gaithersburg, MD), was dried to a constant mass at 50 °C in an oven as per the supplier’s recommendations and was used for method validation. The trial was carried out with different milk samples purchased from a commercial dairy following homogenization with a spatula (for solid samples and SRM infant formula) or by magnetic stirring and heating at 40 °C for 10 min (for liquid samples). For SRM, an accurately weighed amount of 1.0 g was dissolved/diluted in 50 mL of distilled water. For liquid samples, a dilution is required. A 1:10 dilution in water was necessary for cow milk (homogenized and nonhomogenized), a 1:20 dilution was required for goat milk, and 20-25-fold dilution was used for sheep milk. Dilute milk samples were then aspirated into the proposed autoanalyzer (patent granted by University of Co´rdoba, Spain).24 For method validation, samples were homogenized by manual stirring and heated at 40 °C before analysis. Autoanalyzer Functioning. The autoanalyzer designed for milk quality control (Figure 1) operates in a sequential fashion. First, the dilute milk samples are injected (15 µL) in a 50% v/v ethanol-water stream (initially at a flow rate of 0.1 mL/min increasing to 0.8 mL/min after 3 min). In this medium, total protein and fat were precipitated and retained on the cotton filter, with lactose being determined in the filtrate (signal A). In a second step (3 min after sample injection), a pure ethanol stream was passed through the cotton column for 2 min at 0.8 mL/min for (23) McNabb, T. J.; Cremesti, A. E.; Brown, P. R.; Fischl, A. S. Semin. Food Anal. 1999, 276, 242. (24) University of Co´rdoba, Spain, Patent P200202195, September 2002.

1426 Analytical Chemistry, Vol. 75, No. 6, March 15, 2003

Figure 1. Schematic diagram of the autoanalyzer designed for milk quality control. Reagent: 50% (v/v) ethanol-water, pure ethanol, and 1.7 mol/L acetic acid for carbohydrates, fat, and total protein determination, respectively. LPP, low-pressure pump; HPP, highpressure pump; HPIV, high-pressure injection valve; w, waste; ELSD, evaporative light scattering detector. Signals A, B, and C correspond to lactose, fat, and protein detection, respectively.

fat dissolution and direct determination (signal B). Finally, a 1.7 mol/L acetic acid solution (0.8 mL/min, 4 min) hydrolyzed protein (total or partially, as indicated by the ninhidrin test), and the generated amino acids were determined in the ELSD (signal C). Sequential determination of the three parameters was completed within 10 min at room temperature and using a single sample aliquot. Peak height was used as the analytical signal, and the baseline was established with the suitable reagents. After each milk analysis, the flow system was cleaned up for 1 min with an ethanol stream at 0.8 mL/min. RESULTS AND DISCUSSION The major components of milk, apart from water, are carbohydrates, fat, and proteins, the most important protein being casein (80%), and the main carbohydrate, lactose. Proteins can be precipitated using different reagents, such as organic solvents and acetic acid, among others, with carbohydrates (mainly lactose) being determined in the filtrate. The fatty fraction plays a key role in this reaction because it can be either in the precipitate or in the filtrate unless skim milk was used. The fat droplets of the milk are sorrounded by a membrane that consists of phospholipids and a double layer of proteins; membrane proteins are specific in nature. Proteins enter and participate in membrane formation when the fat globule surface area is expanded 4- to 6-fold during the homogenization of milk.25 From this, it follows that under suitable experimental conditions, precipitation/retention of protein/ fat could be achieved, leaving lactose in the liquid phase. Fat is soluble in organic solvents (mainly nonpolar or low-polarity solvents) and insoluble in water. The first assays were developed in glass tubes using 1 mL of milk (whole and defatted) and 1 mL of precipitating reagent, with lactose determined in the filtrate. The best results were obtained with an ethanol-water mixture (similar signals were obtained in both cases). In a second experiment, different organic solvents (e.g., n-hexane, ethyl ether, or ethanol) were assayed for fat extraction from each precipitate (different signals were obtained according to the percentage of fat in both samples). Finally, the remaining precipitate of protein was dissolved with dilute acetic acid, and the signals obtained were comparable for both samples. In all experiments, ELSD was used for detection of the analytical signal. This manual procedure was further automated by using an autoanalyzer similar to that depicted in Figure 1. (25) Belitz, H. D.; Grosch, W. Quı´mica de los alimentos; Editorial Acrı´bia: Zaragoza, Spain, 1997.

Table 1. Figures of Merit of the Proposed Autoanalyzer for the Determination of Lactose, Fat, and Protein in Milk Samples regression eqa

compd

linear detection RSD range (g/L) limit (g/L) (%)

lactose log S ) 0.84, log C ) -0.8 fat log S ) 1.25, log C ) -0.4 total protein log S ) 1.3, log C ) -0.9 a

0.1-12.0 0.1-4.2 0.3-8.2

0.03 0.03 0.08

2.8 3.8 3.2

S ) analytical signal (V); C ) concentration (g/L).

Optimization of Chemical and Flow Variables. The first study was the selection of reagents for protein/fat precipitation/ retention, fat extraction, and protein dissolution. Dilution of the milk samples was necessary prior to introduction into the analyzer in order to achieve good separation between the signals corresponding to each analyte. A 1:10 dilution in water was necessary for cow milk (homogenized and nonhomogenized), a 1:20 dilution was required for goat milk, and 20-25-fold dilution was used for sheep milk. For the protein/fat precipitation/retention, different percentages of ethanol in water (between 10 and 70%, v/v) were tested. The best results were obtained for concentrations higher than 40%, and an ethanol-water solution (50% v/v) was finally selected. It allows complete precipitation/retention of protein and fat on the cotton filter and lactose determination in the filtrate (signal A in Figure 1). For fat extraction, different solvents were tested, namely, n-hexane, ethyl ether, ethanol, and their mixtures. n-Hexane and ethyl ether gave problems associated with their immiscibility in water (from the remaining ethanol-water solution). Unsatisfactory results were also obtained for ethanol-water and ethanol-ethyl ether mixtures, and therefore, pure ethanol was selected for fat extraction (signal B in Figure 1). For the dissolution of proteins, a volatile acid, acetic acid, that allows protein hydrolysis was selected, and its concentration was optimized within the interval 0.5-2.5 mol/L, with the optimum value for protein determination being 1.7 mol/L (signal C in Figure 1). The injected sample volume was also studied. This variable is critical because it should provide the highest sensitivity with the lowest sample dispersion (peak width). Fat extraction and protein dissolution should be easily accomplished. A volume of 15 µL of dilute milk was selected as a compromise. Finally, the flow rates

of the three reagents required for each determination, introduced in a sequential fashion into the autoanalyzer, were optimized by using a high-pressure pump. Lactose determination in the dilute sample requires complete precipitation/retention of protein/fat on the cotton filter with the ethanol-water solution at a low flow rate, 0.1 mL/min, which was further increased (linear gradient) to 0.8 mL/min after 3 min to minimize dispersion of the sample plug containing lactose. The flow rates for fat extractant (pure ethanol) and protein hydrolysis (1.7 mol/L acetic acid) were studied over the range 0.3-1.0 mL/min. Since precipitate dissolution must be instantaneous in order to obtain a transient signal as narrow as possible (with minimal sample dispersion), a flow rate of 0.8 mL/min was selected for both reagents. Sequential introduction of the reagent was finally set as follows: ethanolwater 50% v/v (3 min); pure ethanol (2 min), and acetic acid (4 min). Between samples, the flow system was washed for 1 min with a pure ethanol stream at a flow rate of 0.8 mL/min. Once the autoanalyzer was optimized, several experiments were carried out to corroborate the right peaks assignment. The first peak clearly corresponded to lactose content because it was not precipitated with ethanol-water. The second one was assigned to total fat of the milk, and it was confirmed according to the following assay: One aliquot of whole milk was treated with ethyl ether for fat extraction and then analyzed by using the proposed method. The results were compared with those obtained for whole milk. Negligible differences were obtained for the first and third signals (corresponding to the lactose and protein detection); however, the second one, corresponding to the fatty fraction, practically disappeared in the milk treated with ethyl ether. The third signal corresponded to total protein. Instrumental Conditions. This study was performed by using milk samples with different fat concentrations. Each sample was properly diluted in water, and 15 µL was injected into the autoanalyzer, depicted in Figure 1, working under optimal conditions. The three instrumental parameters affecting the sensitivitys evaporation chamber temperature, nebulizing gas pressure, and photomultiplier gainswere optimized. The evaporation chamber temperature must be selected as a compromise between uniformity of particle size generated and complete solvent evaporation (giving negligible noise) without analyte losses; this parameter was studied in the interval 55-85 °C, and the optimum value was

Table 2. Analysis of Milk Samples by IR (n ) 5) and by the Proposed Autoanalyzer (n ) 15) concn found (g/L) lactose sample 1846a

SRM no. skim cow milk goat milk raw cow milk whole cow milk defatted cow milk raw cow milk skim cow milk raw cow milk sheep milk raw cow milk raw cow milk a

fat

protein

autoanalyzer

IR method

autoanalyzer

IR method

autoanalyzer

IR method

56 ( 2 45.4 ( 0.8 44.7 ( 0.5 50.2 ( 0.6 41.3 ( 0.5 46.3 ( 0.6 50.8 ( 0.6 51.3 ( 0.5 51.0 ( 0.5 45.2 ( 0.5 50.5 ( 0.5 52.2 ( 0.6

57.2 ( 49.1 ( 0.9 46.1 ( 0.6 52.3 ( 0.8 46.7 ( 0.6 48.6 ( 0.7 48.6 ( 0.6 49.4 ( 0.6 47.0 ( 0.6 45.4 ( 0.6 52.3 ( 0.6 48.1 ( 0.7

40 ( 5 3.8 ( 0.1 63.8 ( 0.7 42.0 ( 0.5 33.7 ( 0.5 17.5 ( 0.2 38.5 ( 0.5 4.1 ( 0.1 39.7 ( 0.5 81.1 ( 0.9 37.2 ( 0.4 40.1 ( 0.4

27.1 ( 4.2 ( 0.1 58.2 ( 0.8 42.3 ( 0.6 35.5 ( 0.6 16.6 ( 0.3 40.5 ( 0.5 4.1 ( 0.1 37.6 ( 0.5 76.5 ( 0.9 42.3 ( 0.5 38.8 ( 0.5

12 ( 2 33.8 ( 0.6 32.6 ( 0.4 33.5 ( 0.4 31.9 ( 0.5 36.8 ( 0.6 29.8 ( 0.4 33.2 ( 0.6 28.9 ( 0.5 51.9 ( 0.5 31.8 ( 0.4 35.4 ( 0.4

11.1 ( 0.4b 33.8 ( 0.5 34.2 ( 0.5 31.9 ( 0.5 32.2 ( 0.6 32.9 ( 0.7 33.5 ( 0.4 33.9 ( 0.5 32.2 ( 0.6 55.9 ( 0.6 31.9 ( 0.4 33.2 ( 0.6

1.0b

0.6b

The values of SRM (certified and found) are expressed in percent w/w. b Certified values, not analyzed by the IR method.

Analytical Chemistry, Vol. 75, No. 6, March 15, 2003

1427

65 °C. The nebulizing gas pressure (air flow rate) affects the uniformity and size of the droplets formed and was studied within the range 0.5-2 bar. Peak heights for the three analytes increased up to 1.5 bar, remaining constant over this value. A working pressure of 1.7 bar was therefore selected. The photomultiplier gain was a critical parameter taking into account the variability of the fat content in milk samples (the protein and lactose content are in a narrow interval, regardless of the fat content). Different optimal values were obtained for each analyte according to its concentration in the milk samples (500 V for lactose and 600 V for fat and protein). Finally, the optimal values were selected and the photomultiplier gain was varied during and between analyses. The nebulizer was cleaned monthly by passing an acetone stream through the detector at 2 mL/min for ∼10 min, keeping the air pressure at 2.5 bar and the evaporation chamber temperature at 100 °C. Analytical Performance. The analytical figures of merit for the method were established by using the autoanalyzer depicted in Figure 1. The calibration graphs for the three components studied were constructed by injecting in the autoanalyzer dilute milk samples, previously analyzed by the IR method (to determine lactose, fat, and protein content). The calibration curves were constructed using three replicates for each sample. When the ELSD is used, it was assumed that in a large range of sample sizes, the measured peak area (A) can be related to the samples mass by the following relationship,

A ) amb

where a and b are coefficients that depend on droplet size, concentration and nature of solute, evaporation temperature, etc.26 In the present method, the peak height was selected as the quantitative signal, which was related to the concentration by a double logarithmic expression. The figures of merit for the calibration graphs are summarized in Table 1. Because no blank signal was obtained, the limit of detection was calculated as three times the standard deviation of the peak height for 10 determinations of the same dilute milk at the lowest concentration of the linear range (i.e., 0.1, 0.1, and 0.3 g/L for lactose, fat, and protein, respectively). The detection limits obtained were lower than 0.1 g/L; however, these values are not of relevance, because all milk samples (including skim milk) must be diluted before analysis. The precision, expressed as relative standard deviation, calculated for 11 dilute milk samples (containing 1.0, 0.7, and 0.7 g/L of lactose, fat, and protein, respectively) was acceptable in all instances. Method Validation. To evaluate the performance of the proposed method, the SRM 1846 infant formula was analyzed with the autoanalyzer. Five individual SRM samples were prepared as described in the Experimental Section, and each one was analyzed by triplicate (n ) 15) to determine lactose, fat, and protein. As can be seen in Table 2, the results obtained for lactose and protein were quite consistent. The results for fat were not satisfactory, because they were higher than the certified value. This error (in excess) can be attributted to the breakage of the protein/fat interaction during lyophilization of the SRM. As a consequence, (26) Dreux, M.; Lafosse, M. LC-GC Int. 1997, 10, 382.

1428

Analytical Chemistry, Vol. 75, No. 6, March 15, 2003

Figure 2. Validation of the proposed method by regression analysis against the respective official ones. CA and CIR correspond to the autoanalyzer and IR concentrations, respectively; Se is the estimated standard deviation.

retention of the fat in the filter is no longer associated to the precipitation of casein, because its extraction with ethanol (second step) is faster than that with milk, thus providing a higher and narrower peak. This change in the peak shape makes quantitation impossible using the previously obtained calibration graph based on peak height. A good correlation would have been obtained by using peak areas for calibration, however. An additional test was also performed using three types of milk samples (whole, defatted, and skim) that were fortified with variable amounts of lactose, butter, and bovine casein for lactose, fat, and protein determination, respectively. Average recoveries of 98% for lactose were

obtained for the different fortified samples. Dilute milk samples with variable amounts of butter (80% fat) were spiked, homogenized, and heated at 40 °C; unsatisfactory results were obtained as the likely result of the absence of interaction between the butter and the milk sample, which hindered total retention with protein, which was then partially detected with the lactose fraction (this peak was already wider than expected). Recoveries also failed for protein because of the insolubility of the bovine casein one in the samples, clogging the continuous system and, hence, providing unquantitative results. Considering the difficulties posed by fat determination, the validation studies was finally carried out as follows. Several cow (homogenized and nonhomogenized), goat, and sheep milk samples that were properly diluted to obtain different concentrations of lactose, fat, and total protein were analyzed. Data analysis was carried out using the regression procedure in which the estimated values obtained by the proposed method are represented against the values obtained by the IR method. Figure 2 shows the estimated value vs IR value plot for the three components and the regression parameters for the straight lines obtained. As can be seen, the proposed method performed well as a result of the regression parameters reflected in this Figure. In addition, a paired t test was also used, using 13 samples at varying concentrations for each component; experimental t values were 0.3, 0.8, and 0.5 for lactose, fat and total protein determinations, respectively, being the corresponding critical value of 2.179. No systematic differences were found for all analytes, which corroborates the good performance of the proposed method. Table

2 summarizes the results obtained in the analysis of 11 samples by IR and the proposed autoanalyzer. CONCLUSION The proposed autoanalyzer provides an effective means for the determination of the three macrocomponents (lactose, fat, and total protein) that determine milk quality, which is directly related to its price. The on-line coupled assembly is quite simple and robust; the system offers all the inherent advantages of the automatic methods (e.g., low sample and reagent consumption, minimal manipulation and contact with the reagents, accurate reproducible results and high throughput). In addition, the proposed autoanalyzer is cheaper that those based on IR spectroscopy. However, a high-capacity autosampler and computer-assisted control allow it to work unattended will be required for its successful commercialization. ACKNOWLEDGMENT Financial support from the Spanish DGICyT (Grant BQU20011815) is gratefully acknowledged. The authors thank Prof. M. Jodra´s (Department of Bromatology, University of Co´rdoba) for the valuable discussion of the results. Special thanks is given to M. Herna´ndez from the Laboratory of COVAP commercial dairy (Pozoblanco, Co´rdoba) for her collaboration in method validation and supplying of the milk samples used in this research.. Received for review December 9, 2002.

September

4,

2002.

Accepted

AC020553N

Analytical Chemistry, Vol. 75, No. 6, March 15, 2003

1429